An explicit solution to the matrix equation \(AX - XF = BY\) (Q556908)

The generalized Sylvester matrix equation \[ AX-XF=BY\tag{1} \] is considered, where \(A\in\mathbb{R}^{n\times n},\;B\in\mathbb{R}^{n\times r},\;F\in\mathbb{R}^{p\times p}\). For simplicity, the matrix \(F\) is assumed to be a transposed companion cell. The solution obtained has a neat form, the matrix \(X\) being expressed as the product of the controllability matrix of the pair \((A,B)\), a symmetric matrix associated to the characteristic polynomial of \(A\) and a block Hankel matrix involving \(r\) free vectors. It is shown that if the matrices \(A\) and \(F\) do not have common eigenvalues, the obtained solution represents a complete parametrization of all the solutions of the generalized Sylvester equation. Several equivalent forms of this solution are presented. Some special cases are discussed and the explicit solutions of the (normal) Sylvester equation and of the Lyapunov equation are derived. It is emphasized that the proposed results have important applications in many analysis and design problems in control systems theory. This assertion is proved by considering the parametric pole assignment problem. It is shown that a complete set of solutions of this problem is given by the set of the feedback gain matrices \(K=YX^{-1}\), where \((X,Y)\) is a solution of (1) and \(X\) is nonsingular. The singularity of \(X\) is discussed and two sufficient conditions (\(A\) is cyclic or \(A\) and \(F\) do not have common eigenvalues) are given. Several examples illustrate the proposed solutions.